Ovyka Knowledge Codex

From a search result to the full procedure

A procedure may span a Confluence page, a table, a diagram and a PDF. Ovyka Knowledge Codex (OKC) lets people search the connected content, ask a question and find the page or file that supports the answer. Your teams can then keep reading in the original documentation.

Pages, images and documents in one search

Confluence

Published pages, blog posts and live docs, including sections and tables. Choose the authorised spaces, up to all published content available to the reading account.

Images and attachments

Interface screenshots, business diagrams and images with or without text. Analysis describes what they show and keeps the source page link. Text from supported document attachments is also searchable, with its source page.

PDFs and other files

OKC also indexes PDFs, Word, Excel, PowerPoint and text files in connected file sources and supported Confluence attachments. Search can combine their passages with page results. Ovyka agrees the formats, locations and attachment ingestion path with your team.

You get a common entry point into company knowledge. Each result retains its origin, whether it leads to a Confluence procedure, an image or a document passage.

What becomes searchable in an image

Optical character recognition (OCR) reads visible characters. OKC also analyses the pixels and records the objects, their positions and the visible relationships in words. That AI description is indexed with the surrounding page context and is available for later searches.

For a readable workflow, analysis can capture each step, the named teams, arrow directions, transitions and their conditions. In the “Workflows Jira” example, “incident management” and the reopening rule appear inside the image. Describing them makes it possible to find the page and answer a question about that rule while the page title remains generic.

• A screenshot with text

  An admin screen reads “Pending”. Its visual description records that an hourglass appears on the partner account row. A colleague can search for the label or describe the indicator to find the procedure.

• A diagram with no text

  An architecture drawing can show an application connected to two storage systems, including a fallback route. The description records the visible components and relationships. The page context makes it possible to search for that architecture in the team’s own words.

Details and terms from the analysis enrich the search text. Supported embedded and attached images stay linked to their page and version. The coverage report shows which visuals have descriptions and which still need processing.

What OKC adds for teams using Rovo

Rovo Chat can analyse an image supplied in a conversation. Native processing of images already embedded in Confluence pages remains unavailable to Rovo Chat and its agents according to Atlassian’s open feature request, checked on 6 November 2026. Read Atlassian request ROVO-388.

OKC analyses the attachment during indexing. When someone later describes what they are looking for, the stored description can lead them back to the visual and its source page. They can ask without having a copy of the file. Ovyka can also integrate this index with a Rovo agent via MCP.

How Ovyka implements OKC with your team

Ovyka develops the software and delivers the integration. Our Atlassian and AI specialists work with the people who maintain your Confluence spaces and those who will use the answers.

1. Define the scope.

   We map your spaces, attachment formats and reading permissions with you. You get an agreed list of content to process and the audiences allowed to use it.

2. Index and verify.

   We configure ingestion and visual analysis. Your subject experts ask their own questions, open the retrieved images and check details on the original pages. You get a coverage report and a list of cases to revisit.

3. Connect your tools and hand over.

   We connect the chosen portal or assistant, configure citations and tone, and show your administrators how to monitor updates and retry processing.

Your knowledge in your organisation’s voice

Ovyka works with you to configure the instructions used by the assistant that calls the index. You choose the permitted sources, perspective, brand vocabulary, length and citation format. Consider the same account access procedure in three settings.

Internal knowledge base

“Verify the invitation email address. Point the user to Resend link. Attach the procedure to the ticket.” The technician gets a checklist and an internal Confluence link.

B2B portal

“To renew your access, open the partner portal and enter the address that received the invitation. The Portal access guide explains the remaining steps.” The partner gets a link to the guide they have permission to read.

B2C support

“Has your link expired? Enter your email address on the sign-in page and choose Resend link.” The customer can follow a link to the public help centre.

Connect the index to your AI assistants

Use the OKC web interface or let your existing chat application query federated search. The API and MCP (Model Context Protocol) tools return passages from Confluence or connected files, image descriptions and their sources. Several compatible assistants can use the same tools with their own models. A skill tells them when to search, how to cite evidence and how to format the reply.

Ovyka brings Atlassian expertise, AI engineering and agent integration to that work. An in-house chat, an on-premises assistant, a Rovo agent and an application built on the OpenAI API each have a different integration path. Explore the RAG architecture and connection options.

Product questions

See OKC on a question your team handles

Tell us who needs the information, which Confluence spaces hold it and where they start their search. We will show you the path from question to source page and discuss deployment options for your organisation.